Audibility thresholds of amp and DAC measurements

[Newman]

Hayabusa, thank you! That makes sense as way to produce the number. Might be a wrong calculation, but at least its got some understandable reasoning.

A reason it might be wrong is given in the definition of an absolute threshold of hearing, which is that it only applies to a signal tone in the absence of any other tones. In other words, maybe for a single tone only, the 120dB number might be sensible.

Why are we discussing the -120 number, that I only mentioned in passing as an example? What you asked for, @Wanman, is the meaning of 'strict'. I believe that is answered.

 

[Wanman]

As always, thresholds of audibility are estimates for the average ear, where what is average is represented by the center of a bell curve.

 

[GXAlan]

JSmith, thank you. But there are no references at the end.

That is the primary source

 

[Wanman]

What you asked for, @Wanman, is the meaning of 'strict'. I believe that is answered.

If I understood correctly, It is the result of a string of assumptions by one person, right?

 

[Newman]

Oh come on, that's not the definition/meaning/intent of 'strict', and you know it.

 

[Wanman]

That is the primary source

A little disappointing. I was hoping to for some scientific references.

 

[rwortman]

It’s an audiophile claim that an audio tone that is inaudible by anyone in the absence of all other sounds, somehow becomes audible in the presence of other sounds. There is no sensible reason why this should be so and no demonstration that it is. It assumed because of confounding results of sighted audio ”testing”. It is quite reasonable to assume that a tone that is inaudible when accompanied by dead silence is equally inaudible in the presence of audible sounds. In fact, lossy audio compression schemes rely on the demonstrable fact that tones that are quite audible by themselves become inaudible in the presence of other, louder, sounds.

 

[Wanman]

Oh come on, that's not the definition/meaning/intent of 'strict', and you know it.

Its one person's opinion, at least so far as I can tell.

 

[Wanman]

There is no sensible reason why this should be so...

Hmmm. Are you familiar with stochastic resonance? How about Suprathreshold stochastic resonance?

They are scientific facts in a general way, although one can argue about how they might apply to humans.

 

[GXAlan]

As always, thresholds of audibility are estimates for the average ear, where what is average is represented by the center of a bell curve.

Perhaps, but the standard deviation is unlikely to be greater than 10 dB.

Today’s understanding of exposure to loud noises actually says 110 dB is enough to cause instant damage (again, the frequency matters). But that does mean that the 120 dB is reasonable since 10 dB lower is still a dangerous threshold.

Ref: https://hearing.health.mil/Prevention/Causes-of-Injury/High-Decibel-Levels

The other way to think about is that the OSHA says that 2 minutes at 110 dB can lead to hearing problems and they have a 3 dB exchange rate. So you can do 1 minute at 113, 30s at 116, 15s at 119, 7.5s at 122, 3.75 at 125, 1.875s at 128 dB and then less than one second at 131 dB.

Ref: https://www.cdc.gov/nceh/hearing_lo...at these dB levels,t cause any hearing damage.


Think about oxygen saturation. The average is 97-99% and two standard deviations below gets you to 94%.

Oxygen

Questions and answers about oxygen, oxygen therapy and oxygen saturation (oxygen levels)

www.who.int

So even though there is a bell curve, the standard deviations are small.

 

[Wanman]

...but the standard deviation is unlikely to be greater than 10 dB.

Thank you. That's an interesting claim. Is a gaussian distribution assumed?

Today’s understanding of exposure to loud noises actually says 110 dB is enough...

But what if 110dB is not required?

If there is a detector with a threshold, it is well known that adding noise or another signal can bring the sum of two or more signals above the threshold. In other words, a small signal can ride on a large signal. Some of the time the small signal will then be above the threshold.

 

[Newman]

Hmmm. Are you familiar with stochastic resonance? How about Suprathreshold stochastic resonance?

Only since early last month.

 

[Newman]

Although on reflection, I recall audio discussions from many years ago, mentioning the fact that sometimes we can detect signal below the noise. The question then became how far below the noise? And the answer IIRC was not very far; 3 dB to 6 dB?

 

[GXAlan]

A little disappointing. I was hoping to for some scientific references.

I am confused. Are you suggesting that -120 dB is too strict and the threshold of audibility is closer to practical audibility levels like -90 dB? Or are you suggesting that -120 dB is not strict enough and it has to be something like -130 dB? Or are you suggesting that it has to be -300 dB since one person says that’s the case?

That link describes the methods used in the beginning and he walks through exactly what he is doing. The author is rightfully a primary source.

Noise perception (simulation for white noise)

Simulation of noise perception, detection threshold & dynamic range. SNR & noise audibility (white noise)

www.tonestack.net

If you look at this, you can see what -120 dB might not be good enough and why I like my -120 dB +- 10 dB approach.

I, for one, have stopped chasing transparency and just focus on reducing the circle of confusion.

 

[Wanman]

Hearing a signal below a noise floor is different thing. Apples and oranges.

 

[GXAlan]

Thank you. That's an interesting claim.

The 10 dB is my guess based upon 120 being commonly cited as threshold of damage, 110 being cited by a reputable source that pays Veterans lifelong disability for service connected injuries (the U.S. military) and -130 dB being the theoretical peak from DC-20 KHz based upon that calculator.

You can pick your favorite number if you think I am wrong, but I would love to see what number you think is better?

Is a gaussian distribution assumed?

It doesn’t matter.

See:

http://www2.psychology.uiowa.edu/faculty/mordkoff/GradStats/part%201/I.07%20normal.pdf

But what if 110dB is not required?

If there is a detector with a threshold, it is well known that adding noise or another signal can bring the sum of two or more signals above the threshold. In other words, a small signal can ride on a large signal. Some of the time the small signal will then be above the threshold.

 

[Wanman]

GXAlan, I wasn't trying to declare that I have a better number or anything like that. I was just hoping someone could point me to some published science that supports statements about strict limits in the first post of the thread.

At this point some of the first post still seems like opinion to me. Maybe its opinion based on a string of assumptions.

 

[Newman]

I was just hoping someone could point me to some published science that support statements about strict limits in the first post of the thread.

The funny thing is, it's all in the first post, which has a link to an Amir article (the one I was trying to recall offhand), which in turn has references at the bottom.

The word 'strict' doesn't appear, but the OP of this thread provides his own definition.

So I'm not sure what you are asking for, that isn't already in the material you have already read in this thread.

 

[Wanman]

It doesn’t matter.

On that, I have some concerns. Statistics seems to have gotten more complicated in recent years: https://www.edge.org/response-detail/11715 No longer do bell curves have to be normal.

 

[Wanman]

The funny thing is, it's all in the first post, which has a link to an Amir article (the one I was trying to recall offhand), which in turn has references at the bottom.

The word 'strict' doesn't appear, but the OP of this thread provides his own definition.

So I'm not sure what you are asking for, that isn't already in the material you have already read in this thread.

Newman, yes, I did look at Amir's article that was linked here. IIRC there were two references. I went to AES to read the first page of one of the papers. Didn't see anything about strict limits.

However, IIUC there was a string of assumptions on part of Amir that could be the basis for statements in the first post of this thread about strict limits. In regard to that I think I can see how a number could be gotten if the number applies only to single tones.